Automated detection of amperometric spikes resulting from quantal exocytosis and estimation of spike and pre-spike foot signal parameters

Abstract

Electrochemical microelectrodes can detect single-vesicle release events as "spikes" of amperometric current. We developed a template based "matched-filter" approach that performs least squares fit of a library of templates to the data and identifies a spike when a detection criterion score given by the ratio of amplitude to the standard error exceeds a minimum threshold. This method outperformed existing approaches and detected >95% of true spikes for a mere 2% false positive rate as evidenced by receiver operating characteristic plots of sensitivity vs specificity. The next step is estimation of spike parameters like peak amplitude (Imax), half-maximal width (t50) and area under the curve (Q) which inform maximal flux, flux duration and charge respectively. Closely successive overlapping spikes are ambiguous to estimate as they may not decay back to baseline and should be rejected. Matched filter approach not only provided robust spike detection but also parameter seed values to reject overlapping spikes and also perform iterative curve fitting of spikes. The remaining well-separated spikes were iteratively fit in two phases, first by fitting rising and decaying phases separately and second by fitting the entire time course using seed values from the matched filter template parameters. Using curve-fit parameters, Imax, t50 and Q were calculated. Histograms of these parameters had bi-modal Gaussian distributions with centers and spreads within 12% and 4% of histograms created using manually analyzed data. The pre-spike baseline was estimated using a novel application of the matched-filter criterion scores and the estimation of pre-spike foot signal parameters such as charge (Qfoot) and duration (tfoot) yielded means, and medians within 10% of manually computed parameters.